Fracture　is　a　highly　nonlinear　problem,　especially　under　a　large　deformation　in　cross-linked　elastomers.　Existing　constitutive　theories　and　fracture　models　fail　to　predict　the　fracture　behavior　of　cross-linked　elastomers　because　of　the　absence　of　the　random　polymer　network　structure.　Meanwhile,　few　numerical　methods　can　present　the　realistic　fracture　process　of　a　polymer　network.　In　this　study,　we　propose　a　mesoscopic　network　mechanics　method　to　present　the　entire　large　deformation　and　fracture　process　of　the　polymer　network　in　cross-linked　elastomers.　The　microstructure　of　a　cross-linked　elastomer　is　abstracted　as　a　polymer　network　model　where　nodes　are　referred　to　as　crosslinkers　and　connections　between　nodes　are　referred　to　as　polymer　chains.　A　hybrid　total　energy　form　of　the　network　model　is　proposed　including　the　free　energy　of　polymer　chains　and　volumetric　deformation　energy.　In　addition,　using　a　proposed　stretch　criterion　of　polymer　chains　to　realize　chain　scission,　this　mesoscopic　network　mechanics　method　is　numerically　implemented　by　coding.　Two-dimensional　network　models　with　uniform　and　randomly　distributed　chain　lengths,　as　well　as　models　with　different　types　of　pre-cracks,　are　constructed　and　simulated　under　uniaxial　tension　tests　using　our　method.　Our　method　reproduces　the　entire　hyper-elastic　large　deformation　and　fracture　process　of　polymer　network　models.　The　random　network　models　show　a　lower　Young＇s　modulus　and　a　less　significant　strain-hardening　stage,　indicating　that　the　structural　randomness　is　the　primary　cause　of　accuracy　degradation　for　existing　hyperelastic　constitutive　models　under　large　deformations.　The　results　also　reveal　that　the　structural　randomness　of　the　polymer　network　is　a　primary　reason　for　the　ductile　fracture　and　the　low　notch　sensitivity　of　cross-linked　elastomers.　©　2021